Here are my answers as a client implementor. You'll see our implementation has 
had more time to bake with the application data, but I don't think either is 
technically problematic. I do favor the application data because I believe it 
is better suited for the semantic purpose as a protected result indicator for 
EAP-TLS. When close_notify was floated in the group and put in draft-14, the 
group was very specifically discussing the semantic of "no more messages will 
be sent" - for which close_notify carries that intent very clearly. The group 
has now realized that a protected result indicator is needed instead, and this 
is very different from "no more messages will be sent." It's my opinion that 
close_notify is less suited for this new purpose.

>1. Have you implemented the zero byte application record signa? If not, why 
>not?l
Yes.

>a. Is it sent after receiving the client finished?
Observably in server implementations we are testing against, yes.

>b. Do you anticipate problems if the application record comes before or after 
>a NewSessionTicket message?
No.

>c. did you run into any issues with this mechanism?
No.
>2. Have you implemented the close notify method? If not, why not?
Yes, but testing isn't finished.

>a. Is it sent after receiving the client finished?
We're ironing out a few implementation issues.

>b. Were you able to cause the message to be sent for the server or determine 
>that the message was received for the peer/supplicant?
We're ironing out a few implementation issues.

>c. Did you run into any issues with this mechanism?
Yes, as above - but theoretically I don't see a technical issue once the kinks 
are ironed out. We believe the signal should be accessible to our client 
implementation.

Jorge Vergara
_______________________________________________
Emu mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/emu

Reply via email to